Instead of staying at the table waiting for a mercurial personal computer industry to bring back lady luck, some manufacturers are thinking they'd best leave DRAM for what they see as a safer bet -- made-to-order "foundry" or logic chip making.

Though PCs need ever more memory to run increasingly complex software, keeping some DRAM makers' hopes alive, the switch to foundry chips is alluring. Unlike DRAM, custom-made chips are not a commodity and thus less likely to be vulnerable to a fickle, hyper-competitive market.

"With foundry it's always a pre-agreed price," said Liu Chitung, research chief and technology watcher at SBC Warburg Dillon Read in Taipei. "If you are receiving orders, you know how much money you'll make. It's a better business to be in."

Others say the foundry market is not foreveryone, noting the high technical and marketing demands of custom chipmaking.

Unlike memory chips and computer central processing units such as Intel's Pentium line or the rival PowerPC made by Motorola and IBM, logic chips typically are patented devices designed for a specific purpose. They form the brains or subsystems of digital devices ranging from mobile phones to home electronics to car ignition systems.

One thing is clear: DRAM is no longer the sure thing it seemed to be in 1995 when South Korean memory makers bet big on faster DRAM and bagged $5 billion in profits between them.

Taiwan makers liked that payoff and dozens clamoured to get into the game -- soon contributing to a global glut that since late 1996 has sent DRAM prices freefalling even as demand for wafer making capital and technology has soared.

Experts say Taiwan simply has too many DRAM makers.

"Out of the 40 major factories right now really only 25 are needed," said one analyst, who preferred to remain anonymous.

A yearago, Taiwan's semiconductor firms launched ambitious plans worth some $80 billion over 10 years on new chipmaking capacity. With the global computer slowdown and continuing DRAM price crunch, many firms have scaled back their plans.

Worse yet, Taiwan is in the DRAM game without an ace, since its chip makers lack the research and development prowess needed to squeeze more memory onto a chip -- the only way to compete.

High-end US, Japanese and South Korean DRAM makers now can pack 128 and even 256 megabits on a chip. Taiwan makers, still focussed on 16-megabit chips, must press their Japanese and US allies for licences to still-profitable 64-megabit technology.

Analysts say Taiwan DRAM makers are likely to swallow a loss of some T$26 billion and negative growth of two to eight per cent in 1998 -- far below forecasts of 12-15 per cent growth.

Even 64-megabit DRAM is brutally competitive, down to $10 or $11 a chip for volume orders from $15 in March, when analysts were forecasting end-1998 levels around$13.

So it's no surprise that Taiwan DRAM makers want to muscle into the rich niche forged by Taiwan Semiconductor Manufacturing Corp, the world's leading logic chip foundry.

Though the global slowdown has forced Taiwan's top chip maker to slow its T$400 billion ($11.6 billion), 10-year expansion, Taiwan Semicon's first-half profits nonetheless soared an annual 62 per cent to T$10.7 billion ($310 million).

Analysts say DRAM makers Powerchip, Formosa group unit Nan Ya Technology, Mosel Vitelic Inc are taking custom logic chip orders or plan to start soon.

"Many would like to do it," said Alex Chen, research manager at HSBC Securities Taiwan.

The most conspicuous convert is Acer Inc, which bought out Texas Instruments' stake of their profitless DRAM venture TI-Acer, dubbed it Acer Semiconductor and aims to shift output to "system-on-a-chip" designs in a year or less.

"Getting out of DRAM and into foundry and logic is a crucial strategic move for Acer," said Stan Shih, the flamboyant Acer chairmanwho has taken personal charge of Acer Semiconductor.

Analysts caution that switching from DRAM to foundry is a complex process on technological as well as business grounds.

"The business model and customer base are totally different," said Tsao of ING Baring Securities.

Unlike DRAM, where one maker's chip differs little from another's, foundry work hinges on establishing the flexibility to meet fast-changing market demands and winning the trust of customers whose sales depend on quality microchips.

"With DRAM it's the same thing every day, but with foundry it's something different every day. You have to have a wider range of products and broader capabilities," said Georgiant Ho, senior analyst at Taiwan's Polaris Securities.

"Foundry is more service oriented. Manufacturers can't just walk in and have customers' trust," Ho said.

Shih feels that for Acer the risk is worth taking, noting that the cost of getting into foundry -- estimated by some analysts at T$5 billion ($150 million) -- is comparableto what the firm would lose by staying with DRAM.

"We have a lot of challenges ahead of us, but we have better opportunities than others," he said. "With DRAM we would have lost the same amount, so it's a good time for us to make a move."

Hope is not lost for those who choose not to abandon DRAM, though most analysts see commodity memory as a long-shot gamble for at least another year even if a manufacturer can stay at the front of the technology curve.

"It's a cyclical business, but I don't think we will see any kind of a recovery until late 1999," said Liu of SBC Warburg.